The good news is we can use the Keras API for training deep models on Google Colaboratory TPU. But there are several important things we need to know in order to do so.
This post is a slightly modified chapter from my ebook “Machine Learning with Clustering“
There’ve been proposed several types of ANNs with numerous different implementations for clustering tasks. Most of these neural networks apply so-called competitive learning rather than error-correction learning as most other types of neural networks do. ANNs used for clustering do not utilize the gradient descent algorithm.
Probably, the most popular type of neural nets used for clustering is called a Kohonen network, named after a prominent Finnish researcher Teuvo Kohonen.
There are many different types of Kohonen networks. These neural networks are very different from most types of neural networks used for supervised tasks. Kohonen networks consist of only two layers.